How do you keep an AI chatbot or receptionist up to date as your business changes?
Keep an AI chatbot or receptionist up to date with a routine. Name one person who owns the facts it uses. Decide which changes trigger an update, such as prices shown, hours, areas, policies, staff, seasons or a new service. Each month, review what it was asked and could not answer. Retest a fixed set of real questions after every change, remove old versions of documents, and keep a dated change log.
One owner, and what should trigger an update
Give one person the job of keeping the assistant's facts right, and name a backup. If everyone can change them, often no one does. The owner is the person who hears about a change first, which in a small business is often you.
Then decide which changes count. Any of these can leave the assistant giving last month's answer in your name:
- Prices you show customers, or how you quote.
- Hours, including holidays and seasonal hours.
- Service areas, and the jobs you now take or no longer take.
- Policies, such as cancellations, deposits and refunds.
- Staff changes, including who is on call and who answers which question.
- Seasonal offers that start or end.
- A new service, a new phone number or a new booking link.
Make it part of the change, not an extra job
The easiest habit is to add "update the assistant" as the last line of whatever you already do when one of those things changes. Update it the same day.
A short monthly review
Once a month, read the questions customers asked that it could not answer or passed to a person, if your tool keeps them. Sort each into one of three piles: a fact it was missing, a topic that should stay with a person, or a question not worth answering. Repeated questions are often the best list of what to add next.
Retest and keep a change log
After each change, rerun your fixed set of real customer questions, and remove the old version of any document you replace; the related answers on stopping an AI assistant from making things up and on helping train new staff cover the test set and keeping documents current.
Keep a dated change log: what changed, who changed it and what the retest showed. It can help you connect a wrong answer to a recent change; the related answer on what to do when an assistant gives wrong information covers finding the cause.
When a tool it relies on changes
An assistant usually rests on things you do not control: the vendor's software, the model underneath it, and the calendar or lead list it connects to. Vendors can change features or plans, and a changed password or a moved account can break a connection. Ask your vendor how it announces changes, read those notices, and rerun your test set when one arrives or when answers start to read differently. The related answer on monitoring an AI system after launch covers watching that the connections keep working.
Related answers
- How do you stop an AI assistant from making things up?
- Can an AI assistant help train new staff?
- What should you do when your AI assistant gives a customer wrong information?
- How do you keep an eye on an AI system once it is running?
AI assistants and chatbots
A good assistant answers the question a customer asks at nine at night, in your words, from your information, and brings in a person when it should. We build that, and we test it on your real questions before a customer sees it.